Vessel Segmentation And Get Rid Of Calcified Plaques In CT Images Of Coronary Vessels

The mortality of coronary heart disease, a universal disease, stay in a high level stubbornly. To analyse the severity of coronary artery stenosis,researchers all over the world employed FFR, which stands for Fraction Flow Reserve, for its clinical significance. However, precise 3D model of patient' s coronary artery is needed before the calculation of FFR. This paper is to introduce artery segment as well as the detection and removement of calcified plaque in the model construction.Traditional image segment algorithms can be generalized into two series, the ones focus on edges and other ones on regions. The former detect the vagaries of eigenvalue in neighbor pixels to figure out and connect the boundaries to separate images into different regions, like the edge detecting technique based on derivative operators. Nevertheless, the later merge pixels or smaller regions which resemble in local feature like gray scale or grain into greater region. For example, region growth algorithm, watershed algorithm, etc.However, vessel crisscross around cardiac distribute in a complex way there. Additionally, difference of images' pixel values appear similarly so that traditional segment algorithm can not work out its way processing them. We explore traditional algorithm in depth with examing their strength as well as weaknesses, combine the features of coronary artery with its branches and finally introduce a 3D image segment algorithm,specificly for coronary artery, in this paper. As proved by experiments, the algorithm works out more efficiently and precisely way.Moreover, the calcified plaque attached to walls of coronary artery, if not removed in advance, could change blood flow velocity, causing great influence on calculation precision of FFR. Therefore, removement of calcified plaque in CT images is in great necessity. These are main steps taken to detect and remove calcified plaque. Firstly, this paper employs HOG and LBP features with SVM for initial detection to avoid mistake.Then, we introduce directional expansion algorithm and plaque' s outline fitting algorithm based on center line of artery. After the comparison and analysis with experiments, methods proposed in this paper reach a good precision in detection and removement of calcified plaque.